UCLA

This thesis investigates and demonstrates observational transportability, a causal inference approach that combines observational information from a study population and observational covariate data from a target population to generate potential causal insights in the target population. Inspired by transportability in the setting of a randomized control trial (RCT), a set of identifiability assumptions for observational transportability is provided. Then, a general formula is obtained for the average causal effect in the target population (TACE) under this framework. The concept is then used to extrapolate the average causal effect of blood lead on hypertension from a study population represented by the National Health and Nutrition Examination Survey (NHANES) to a target population represented by the Behavioral Risk Factor Surveillance System (BRFSS), thereby computing the average causal effect in the target population. Limitations of observational transportability in generating unbiased causal effects in target populations are discussed. The study concludes by considering observational transportability in practice, including an approach enabled by extracting models published in other studies.